With Python 3 build, do not install python2

[jhpalmieri]

When Sage is built with Python 3, there is no need to build Python 2, so let's skip it.

CC: @jdemeyer @fchapoton @embray @kiwifb @slel @antonio-rojas @isuruf

Component: python3

Author: John Palmieri

Branch/Commit: fa1482d

Reviewer: Frédéric Chapoton

Issue created by migration from https://trac.sagemath.org/ticket/28426

[jhpalmieri]

Branch: u/jhpalmieri/no-python2

[jhpalmieri]

Commit: f5de6f9

[jhpalmieri]

comment:2

With this branch and a Python 3 build, I get two new doctest failures:

sage -t --warn-long 60.3 src/sage/doctest/control.py
**********************************************************************
File "src/sage/doctest/control.py", line 630, in sage.doctest.control.DocTestController.test_safe_directory
Failed example:
    DC.test_safe_directory(d)
Expected:
    Traceback (most recent call last):
    ...
    RuntimeError: refusing to run doctests...
Got:
    <BLANKLINE>
**********************************************************************
1 item had failures:
   1 of   9 in sage.doctest.control.DocTestController.test_safe_directory
    [204 tests, 1 failure, 4.24 s]
sage -t --warn-long 60.3 src/sage/tests/cmdline.py
**********************************************************************
File "src/sage/tests/cmdline.py", line 412, in sage.tests.cmdline.test_executable
Failed example:
    print(err)
Expected:
    ...
    RuntimeError: refusing to run doctests...
Got:
    <BLANKLINE>
**********************************************************************
File "src/sage/tests/cmdline.py", line 416, in sage.tests.cmdline.test_executable
Failed example:
    print(err)
Expected:
    ...
    RuntimeError: refusing to run doctests...
Got:
    <BLANKLINE>
**********************************************************************
1 item had failures:
   2 of 245 in sage.tests.cmdline.test_executable
    [244 tests, 2 failures, 69.42 s]
----------------------------------------------------------------------
sage -t --warn-long 60.3 src/sage/doctest/control.py  # 1 doctest failed
sage -t --warn-long 60.3 src/sage/tests/cmdline.py  # 2 doctests failed
----------------------------------------------------------------------

Should these be marked # py2?

---

New commits:

f5de6f9

trac 28426: In a Python 3 build of Sage, do not build Python 2.

[fchapoton]

comment:4

see the discussion in #27529 about the same failing doctests

[jhpalmieri]

comment:5

Replying to @fchapoton:

see the discussion in #27529 about the same failing doctests

One comment there was that if python is python3, then the tests would work. I just created a symlink python -> python3 here, and the tests still fail. I think either marking the tests as py2, or removing them (#26457) would be good. Or would #25358 help?

[fchapoton]

comment:6

Let us just tag them with #py2.

[sagetrac-git]

Changed commit from f5de6f9 to fa1482d

[sagetrac-git]

Branch pushed to git repo; I updated commit sha1. This was a forced push. New commits:

656f6b7	trac 28426: In a Python 3 build of Sage, do not build Python 2.
fa1482d	trac 28426: tag some doctests as 'py2'

[jhpalmieri]

comment:8

Okay, ready for review.

[fchapoton]

comment:9

This looks good to me. I would like to have more opinions from experts, in particular people packaging sage for various Linux distributions.

And this will have to wait for 9.0.beta1, I think.

[embray]

comment:10

I agree this should wait for 9.0 at the earliest, but +1 in principle. No reason to burden people with two Pythons.

[fchapoton]

comment:11

So can we switch to positive review ? No packager did react.

[davidlowryduda]

comment:12

For what it's worth, this also looks good to me and works on a minimal arch setup and an ubuntu 19.04 setup.

[fchapoton]

comment:13

I am setting to positive, for entry in 9.0.beta0

[fchapoton]

Reviewer: Frédéric Chapoton

[jhpalmieri]

comment:14

I'm wondering about a followup ticket, but I don't know if this would be a good idea: currently, the script sage-python23 checks for the value of SAGE_PYTHON3 to decide which version of Python to run, and the same is true when running sage --python. Should we (a) actively delete local/bin/python when building with Python 3 (relevant in the case of updating existing installations of Sage), and then (b) have sage-python23 and sage --python run SAGE_LOCAL/bin/python if if exists, SAGE_LOCAL/bin/python3 otherwise? Is it good to not depend on the environment variable SAGE_PYTHON3?

If we ever hope to use a system version of Python, I guess this would be a bad idea.

[vbraun]

Changed branch from u/jhpalmieri/no-python2 to fa1482d